The proof deck

Who shipped what—and what happened next.

This is not a member directory. Every entry starts with a live artifact, a declared threshold, and evidence reviewed against the original Bet.

No Ship in 30 result has passed verification yet. Reference records below show the evidence standard without pretending to be participant proof.Take on Ship in 30 →

Annotated reference records

See the decision change.

These internal Product Builders records are teaching examples, not verified challenge entries. They show the original Bet, the evidence that forced a revision, what can honestly be claimed, and what remains unknown.

Reference proof 01 · Practice productShipped

Turn a discovery lesson into a runnable field kit

Should Product Discovery remain an explanatory lesson or become a workflow a builder can run end to end?

Original Bet: A staged five-day workbench will be more useful than another page of methods, questions, and rubric copy.
Threshold: One saved flow must frame the decision, recruit, capture interviews, synthesize evidence, force a call, and export the memo.
Current result: The static lesson was replaced by a 15-action workbench with evidence gates, local persistence, memo export, and a reusable agent skill.
Next move: Measure sprint starts, completed evidence records, exported memos, and return-to-finish behaviour before claiming learning impact.

Product DiscoveryUser ResearchWorkflow Design

Run the field kit →Inspect the skill bundle →

Evidence ledger

Observation → consequence

Problem
The page described discovery but gave the builder nowhere to conduct it.
Replace passive instruction with a staged working surface.
Revision
A method, question list, exercise, and rubric still left the hard operational steps implicit.
Add recruiting, structured evidence, decision gates, and required artifacts.
Current evidence
The complete workflow is usable and testable; real completion and outcome data is not yet available.
Keep the claim at “runnable,” instrument use, and do not call the skill mastered from checkboxes.

Reference proof 02 · Editorial systemChanged course

Stop publishing the same brief under a new date

Should a daily slot be filled when the available signal is stale or semantically repetitive?

Original Bet: A trustworthy gap is more valuable than a manufactured daily edition.
Threshold: A candidate needs a fresh dated source, a new source URL, and no duplicate title, summary, thesis, or recent topic pattern.
Current result: Exact duplicate editions were removed and the generator now refuses weak or stale signals instead of silently cloning the last thesis.
Next move: Expand the guardrail from exact duplication to discipline coverage, semantic topic cooldowns, primary-source strength, and follow-up reporting.

Editorial JudgmentProduct AnalyticsSystem Quality

Read the current archive →Open today’s briefing →

Evidence ledger

Observation → consequence

Signal
Three consecutive archive entries carried the same title, summary, and practice instruction.
Treat recurrence as a pipeline failure, not a presentation bug.
Gate
Date-based IDs prevented collisions but said nothing about freshness or distinctiveness.
Check the source and editorial substance before serialization.
Current evidence
The duplicate case is covered; the archive remains too concentrated on agents and workflows.
Do not confuse a fixed defect with a sufficiently broad intelligence product.

Reference proof 03 · InterfaceShipped

Make the product-builder model inspectable, not decorative

How should the homepage explain the collapsed product trio without another static Venn diagram?

Original Bet: A stage-based instrument can demonstrate inputs, current judgment, and proof artifacts more clearly than overlapping circles.
Threshold: The model must expose all five stages, respond to direct selection, pause, respect reduced motion, and connect to a current signal.
Current result: The static graphic became a selectable Sense–Shape–Build–Ship–Learn instrument with live brief context and explicit outputs.
Next move: Turn the explainer into a diagnostic: ask where the builder is stuck, recommend the correct lab, and prefill the first task.

Information DesignUI DesignProduct Narrative

Inspect the homepage instrument →Start from Today instead →

Evidence ledger

Observation → consequence

Problem
The earlier visual made “product, design, engineering” look like a branding diagram rather than a working model.
Show the sequence of judgment and the artifact produced at each stage.
Revision
Automatic rotation alone would still make the component decorative.
Add direct stage controls, pause, state labels, and keyboard-accessible buttons.
Current evidence
The model is understandable and interactive but does not yet accept the builder’s own situation.
The next version must start work, not only explain the brand.